An autonomous biomimetic walking kangaroo robot
Tasked with four weeks to design, engineer, and build a mechanical robot that mimicked the walking kangaroo, my team—four of us—set to applying our knowledge of linkage mechanisms, power flow, motor characterization, and gear theory.
Make a robot that mimics a walking kangaroo as closely as possible.
The kangaroo robot must:
- travel across laid brick at about ten centimeters per second.
- closely mimic the pentapedal locomotion of a walking kangaroo (even proportion of work done by each limb).
- be autonomous and battery-powered.
- be no longer than thirty-five centimeters in any dimension.
- cost under eighty dollars.
Because we had become linkage geeks, we explored the implementation of complex linkages—Klann, Hoeken, and Jansen—but quickly foresaw possible weight issues; our electric motor would have drawn too much current. Moreover, these linkages' strengths—a rapid return stroke and human-like curve paths—did not fit the task hand.
Instead, we opted for the simpler pin-in-slot and four-bar linkages. For the hindlimb and tail, we selected the pin-in-slot mechanism for its tendency to generate quicker return strokes, but still maintain a flat and constant-velocity contact with the ground. Because of four-bar linkages’ versatility in application, we designed our forelimb with the mechanism.
I led the linkage modeling process. We first used visual online software to approximate linkage curves; then, building from Matlab functions, I wrote Python scripts that detailed the curves and confirmed the feasibility of the design.
We constructed a prototype of the four-bar linkage to validate our models.
While researching the movement of kangaroos, we noticed two important features that guided our design: the paths of the forelimbs and hindlimbs overlapped and the kangaroo’s forelimbs almost exclusively do resistive—or bracing—horizontal work.
To mimic the overlapping legs, we placed the legs within stroke distance of each other and used ¼” acrylic spacers between the body of the Kangaroo and its legs; this extra spacer placed the hindlimbs in a parallel plane to the forelimbs with half an inch of clearance. Additionally, we built the tail limb to within the body, such that it more precisely looked like a kangaroo and did not interfere with the hindlimbs.
Weight Distribution key to biomimicry
To appeal to a real kangaroo’s walking motion, we had to carefully manipulate the balance of our robot by adjusting the angle of its body, the center of mass, and the phase of its legs. In resting position, the front legs are short compared to the rear legs, and the tail is long compared to both.
The relative lengths leans the robot forward, mimicking the kangaroo's walking posture, but placing undue weight on the front legs. By placing the battery pack in the rear, the center of mass sat directly over the rear legs.
We used a Tamiya 72005 6-Speed Motor with a 196.7:1 speed reduction, the minimum step that would provide enough torque to propel the kangaroo. Due to friction throughout the kangaroo, the motor drew an average of approximately one ampere, which caused our motor to operate below the nominal three volts; on average, the motor operated at about two volts.
The motor's drive axle mechanically attached to the two hindlimb axles via pin—direct drive style—to deliver power to the hindlimb axles, which then transmitted power to the tail and forelimb axle. A chain and sprocket connected a hindlimb axle to the tail axle; a belt-pulley system transmitted power to the forelimb axle.
Power Loss Reduction
We could not affect power loss within the motor or transmission and instead focused on reduction of loss due to friction and slippage. Thus, we implemented lubricated bushings, hand-adjusted joints for a minimum acceptable friction (qualitatively), and attached sand paper to the robot's feet.
Analysis & Redesign Plan
The belt-pulley system required a small range of tension to function properly. Tolerance stack in manufacturing caused the insufficient tension to be applied to the belt, resulting in occasional slippage (seen in the slo-motion video, left). Belt slippage led to accumulating misalignment between limb phasing. As an ad hoc measure, we implemented an idler, which both increased tension and wrap angle.
We bypassed three-dimensional modeling in our housing design; while putting together our laser-cut parts, we realized the cost cutting this corner: our design intersected belts and sprockets with components of the housing. To accommodate, we bandsawed housing parts, but the result was less stable and less clean.
Due to an over-tensioned belt and excess friction in joints, the Kangabruiser would operate at an insufficient motor RPM, not achieve a distance of ten centimeters per second. The robot also traveled at an angle slightly to the left. Lower and more even friction would reduce wear and increase performance simultaneously.
For further details and an in-depth analysis of performance, see The Kangabruiser Design Report. In addition to leading locomotion modeling and contributing to many aspects of design and the build, I wrote the Design and Redesign sections of the analysis, and led editing efforts.